嵌入式固件可靠性审计
围绕掉电、中断、内存和外设异常检查固件恢复能力,输出故障证据、修复建议与故障注入验证方案。
完整提示词
I want a deep reliability and correctness audit of the embedded/firmware system with a focus on timing, interrupts, memory, persistent state, the watchdog, power loss, hardware interaction and recovery. Main objective: > Determine whether the firmware can safely continue operating or recover when a reset, power loss, noisy input, partial peripheral failure, timing pressure, communication corruption or memory exhaustion occurs. This is not: - a generic C/C++ review - a hardware schematic audit - RTOS evangelism - an automatic recommendation of a dynamic or static memory model - a safety certification If the system is safety-critical: clearly state that this audit is not a substitute for formal certification. ## 1. PLATFORM - MCU/SoC - architecture - clock - RAM - flash - peripherals - RTOS/bare metal - bootloader ## 2. RESET PATH ## 3. POWER-ON ## 4. BROWNOUT ## 5. WATCHDOG ## 6. WATCHDOG FEED ## 7. WATCHDOG BLIND SPOT Task hangs but another task feeds watchdog. ## 8. BOOTLOADER ## 9. FIRMWARE UPDATE ## 10. A/B SLOT ## 11. ROLLBACK ## 12. FAILED UPDATE ## 13. IMAGE SIGNATURE ## 14. VERSION ## 15. DOWNGRADE ## 16. INTERRUPT ## 17. ISR LENGTH ## 18. PRIORITY ## 19. SHARED DATA ## 20. VOLATILE Not concurrency primitive. ## 21. ATOMIC ## 22. CRITICAL SECTION ## 23. INTERRUPT MASKING ## 24. DMA ## 25. CACHE COHERENCY Where relevant. ## 26. RTOS TASK ## 27. PRIORITY INVERSION ## 28. DEADLOCK ## 29. STARVATION ## 30. SCHEDULING ## 31. DEADLINE ## 32. JITTER ## 33. WCET Where needed. ## 34. TIMER ## 35. TICK WRAPAROUND ## 36. INTEGER WRAP ## 37. CLOCK ## 38. RTC ## 39. CLOCK DRIFT ## 40. COMMUNICATION - UART - SPI - I2C - CAN - USB - BLE - Ethernet ## 41. FRAME VALIDATION ## 42. CRC ## 43. TIMEOUT ## 44. RETRY ## 45. BUS LOCK ## 46. DEVICE ABSENT ## 47. PARTIAL PERIPHERAL FAILURE ## 48. SENSOR ## 49. OUT-OF-RANGE ## 50. NOISE ## 51. DEBOUNCE ## 52. ACTUATOR ## 53. SAFE STATE ## 54. FAIL-SAFE ## 55. FAIL-OPERATIONAL If requirement. ## 56. MEMORY ## 57. STACK ## 58. STACK OVERFLOW ## 59. HEAP ## 60. FRAGMENTATION ## 61. BUFFER ## 62. BOUNDS ## 63. USE-AFTER-FREE ## 64. STATIC LIFETIME ## 65. RESOURCE HANDLE ## 66. FLASH ## 67. EEPROM/NVM ## 68. WEAR ## 69. ATOMIC PERSISTENCE ## 70. POWER LOSS DURING WRITE ## 71. CRC/VERSION OF STORED DATA ## 72. FACTORY RESET ## 73. CALIBRATION ## 74. CONFIG ## 75. CORRUPT CONFIG ## 76. LOGGING ## 77. RING BUFFER ## 78. LOG STORM ## 79. STORAGE FULL ## 80. DIAGNOSTICS ## 81. SAFE MODE ## 82. RECOVERY MODE ## 83. ASSERT ## 84. PRODUCTION ASSERTION ## 85. HARD FAULT ## 86. CRASH DUMP ## 87. RESET REASON ## 88. BREADCRUMB ## 89. POWER ## 90. SLEEP ## 91. WAKE ## 92. RACE AFTER WAKE ## 93. LOW VOLTAGE ## 94. THERMAL ## 95. EMI Software-level resilience only. ## 96. TEST ## 97. HIL ## 98. SIL ## 99. FAULT INJECTION ## 100. FUZZ Protocols/parsers. ## 101. LONG-RUN ## 102. RESET LOOP ## 103. FALSE POSITIVE RULES Do not automatically flag: - bare metal - RTOS - dynamic allocation - interrupts - polling - watchdog without actual reliability consequence. ## 104. EVIDENCE TIERS ```text A - reproduced HIL/runtime/fault-injection failure B - complete timing/state/hardware interaction proof C - strong static evidence D - plausible embedded failure E - hardening ``` ## 105. STATUS ```text CONFIRMED LIKELY NOT VERIFIED CONTROLLED NOT APPLICABLE HARDENING ``` ## 106. SEVERITY P0: unsafe uncontrolled hardware behavior, systemic bricking or unrecoverable critical state P1: repeatable severe reliability/persistent-state/watchdog/update defect P2: material timing/peripheral/recovery issue P3: limited robustness issue P4: hardening ## 107. FINDING FORMAT ```text ID: Severity: Status: Evidence tier: Hardware: Firmware version: Task/ISR: Peripheral: Trigger: Timing/state sequence: Expected invariant: Actual behavior: Safe state: Recovery: Evidence: Root cause: Fix: HIL/fault regression: ``` ## 108. FAILURE MATRIX | Failure | Detection | Safe state | Recovery | Persistent impact | |---|---|---|---|---| ## 109. SECOND PASS Inject: - power loss during write - peripheral absent - corrupt packet - stuck bus - high interrupt load - task starvation - low memory - watchdog reset - update interrupted - corrupt config - timer wrap - repeated sleep/wake ## 110. FINAL QUALITY GATE Confirm: - reset - boot - update - interrupts - RTOS - timing - communication - peripherals - memory - persistent data - watchdog - safe state - power - diagnostics - fault injection ## 111. OUTPUT `EMBEDDED_SOFTWARE_RELIABILITY_AUDIT.md` ## 112. FAILURE CHAIN ```text configuration written in-place to flash ↓ power fails halfway ↓ CRC not stored separately ↓ next boot reads partially updated configuration ↓ actuator starts with invalid threshold ``` # FINAL RULE An embedded reliability audit must always ask: > What happens when the system loses power, time, communication or a peripheral in the middle of a state transition?
分类与标签
编程开发、嵌入式固件、掉电恢复、中断与时序、看门狗、故障注入